1. Introduction
Afghanistan is frequently ranked among the countries most vulnerable to climate change (e.g., [
1,
2,
3]) due to a combination of low adaptive capacity and high exposure to climate fluctuations. Over the past four decades, armed conflict has destroyed the country’s infrastructure, damaged its institutions, and led to widespread poverty and underdevelopment, which collectively underpin Afghanistan’s vulnerability and lack of adaptive capacity to climate change [
4]. The population and the economy are almost completely dependent on agricultural production, particularly subsistence farming [
5], and key sectors, including water, energy, agriculture, are among the most vulnerable to climate change. The country is regularly hit by extreme weather or climatic events, causing substantial economic damage and loss of lives [
1,
6], showing that even today Afghanistan is not sufficiently adapted to the current climate.
Despite this very alarming situation, almost no scientific literature on climate change and its impacts, in the past nor projected for the future, exists [
7]. Most adaptation initiatives rely on one study of the Tyndell Centre for Climate Change Research done for the United Nations Development Programme, where global climate models of the third phase of the Coupled Model Intercomparison Project (CMIP3) were analyzed for Afghanistan amongst all other nations of the world in 2010 [
8]. The study found that mean annual temperature has increased by 0.6 °C since 1960 in Afghanistan, while mean rainfall decreased slightly. In this study, the analysis of the past is based on station data which have large gaps, especially after 1970. In addition, the explanatory power of the projections is limited due the coarse resolution of the CMIP3 global models of 2.5° (Afghanistan is covered by 19 cells), particularly given the extremely mountainous character of the country. Another study by the Stockholm Environment Institute used the same data sets to analyze the socio-economic impacts of climate change on Afghanistan [
9], finding that the “adaptation challenges facing Afghanistan are very significant in scope and scale.” Another study by Ridley et al. (2013) [
10] found that the Karakorum will receive more precipitation due to an increase of westerly disturbances. This is confirmed by the study by Mukhopadhyay and Khan (2014), which projects a warming of 2 °C and a slight increase of precipitation (8–10%) until 2050 for the Upper Indus Basin, including its Hindukush part [
11]. In addition to the lack of literature on climate change, there are also very few impact studies for Afghanistan. For glaciers in the Pamir/Hindu Kush region and their influence on the water resources of the Amu Darya basin, of which Afghanistan has shares in the upper catchment, several climate impact studies exist for past [
12,
13,
14,
15] and future projections [
16,
17,
18,
19,
20,
21]. Other relevant studies exist on the general influence of climate change and human impact on water resources in the Amu Darya basin, including the Aral Sea as part of this basin [
22,
23,
24]. Other studies focus on hazards, mainly floods [
25,
26] and landslides [
27], but without a climate change perspective. Beside these studies, some grey literature on climate change effects in Afghanistan exists that is mostly not based on systematic scientific approaches since the reports in the development context have different scopes (e.g., [
4,
28,
29,
30]). The main reason for this large scientific gap is most likely the difficult working conditions during the long period of conflict and the lack of data and capacity in the country. Moreover there is a strong focus on peace and development of the international community rather than scientific research. In addition, the extreme heterogeneous geography of Afghanistan, ranging from the glacier-covered Hindu Kush in the north to the arid deserts in the south, impede transferring larger-scale climate change studies to Afghanistan.
The present study intends to start filling this gap and is the first analysis of past and future climate trends using the outcome of regional climate models with explicit focus on Afghanistan. The goal is a solid analysis of climate change for both the past, using reanalysis data due to the limited observed data, and future, using an ensemble of state of the art regional climate models from the Coordinated Regional Climate Downscaling Experiment (CORDEX) for South Asia, which includes an assessment of data uncertainty.
As the scientific literature on Afghanistan’s climate is limited and Afghanistan’s geography might not be known to all readers, the sub-regions of Afghanistan are briefly described in the following section, with a focus on the climatically relevant context. The reanalysis data is validated against the available station data and, subsequently, the ability of the models to reproduce the current climate of Afghanistan is evaluated in order to determine systematic biases and to assess the confidence we can have in their future projections. With the reanalysis data, differences between the 30-year periods of 1981–2010 and 1950–1980, as well as linear trends over the whole period from 1951 to 2010, are analyzed.
In addition to the standard parameters for annual temperature and annual precipitation, five additional climate indices for extremes and agriculture assessments are applied, which are especially relevant for Afghanistan’s most common hazards and its agricultural system (heavy precipitation; spring precipitation; growing season length (GSL), the Heat Wave Magnitude Index (HWMI); and the Standardized Precipitation Evapotranspiration Index (SPEI)). The same parameters are subsequently analyzed for the future model projections. The difference between the future period 2021–2050 and the base period 1986–2005 is calculated and mapped in order to visualize the regional patterns of the expected changes for an adaptation-relevant, near-future scenario. Linear trends are analyzed, starting in 2006 when the scenario data of the climate models begin again, until 2050. The trends are additionally analyzed until 2099 in order to show long-term changes. In order to cover a broad range of potential future scenarios, a low emission (Representative Concentration Pathway (RCP) 4.5) and a high emission scenario (RCP 8.5) are used for the projections. Finally, the results and their uncertainties are discussed, including not only the pure climatic changes in the past and future but also their implications for the agricultural sector and water resources since these areas are considered crucial to national development.
3. Results
3.1. Validation of Reanalysis
The validation of the reanalysis and historical model data is done based on monthly time series and in a qualitative way (
Figure 3 and
Figure 4); the results and their implications on uncertainty are discussed in
Section 4.1. The seasonal variations of temperatures of the reanalysis data are in good agreement with meteorological measurements at all stations. The absolute magnitudes of temperatures are well reflected for Gardez (East), Herat (Northwest), Kandahar (South), and mostly for Panjab (Center). For Faizabad (Hindu Kush), Mazar-e Sharif (North), Kabul and partly Panjab (Centre) the absolute magnitudes differ. However, for Faizabad, the reanalyzed and monthly averaged temperatures are underestimated by approximately 5 °C and for Mazar-e Sharif and Kabul by approximately 3 °C to 5 °C. For Panjab, the winter temperatures are underestimated by approximately 5 °C. These differences are discussed in
Section 4.1.
The seasonal cycle for precipitation is generally correct for all stations with one rainy season in the North during winter and spring and for the stations in the East that are influenced by the summer monsoon, an additional small rainy season occurs in summer (Gardez, Kabul, Kandahar). In terms of absolute monthly magnitudes, Herat (North) and Kandahar (South) are well in line with observations during all months. For the other five stations, monthly precipitation between May and December is reasonably estimated. Precipitation between January and April for these stations is, however, mainly overestimated by up to approximately 40 mm in some cases.
3.2. Performance of Climate Models
The comparison between the historical climate model runs and observations shows similar patterns for temperature even though for some stations the absolute temperature level differs. The models are able to reproduce the seasonal temperature cycle for all stations. The absolute values of model mean and the simulated range of monthly temperature are in the same ranges as in the observations for all stations except for Faizabad, Mazar-e Sharif and Kabul. Here the deviations range in the same order of magnitude as in the reanalysis. This is further discussed in
Section 4.1.
With respect to precipitation, the results differ completely. Only for stations Faizabad, Mazar-e Sharif and partly for Herat in the northern part of Afghanistan are the models capable of reproducing the observed seasonal cycle of monthly precipitation. For all the other stations, the models particularly underestimate the spring precipitation from January to April. Only the onset of winter precipitation is simulated correctly for these stations; however, not the observed magnitudes. This has strong implications on the interpretability of these data which is discussed in detail in
Section 4.1.
3.3. Analysis of Past Climate Trends for the Period 1951–2010
The general climatic trends are depicted in the
Figure 5a,b and quantified for the regions in
Table 4. In the reanalysis, the mean annual temperature increased substantially in most parts of Afghanistan by up to 1.2 °C in a diagonal northwest–southeast belt west of the Central Highlands for the difference between 1981/2010 and 1951/1980. The temperature for the whole period from 1951 to 2010 for all Afghanistan even increased by 1.8 °C, with highest increases of 2.4 °C in the east and only 0.6 °C in the Hindu Kush region. Towards the northeast and west, the temperature increase fades. No increase is detectable in the region southeast of Kabul. In Badakhshan, in the northeast, where Afghanistan’s main glaciered areas are located, the warming is between 0.3 and 0.7 °C. All these trends are statistically significant.
The trends in the reanalysis for annual precipitation are less distinct for most parts. In addition they are not significant for all regions. Most parts of Afghanistan experienced changes between −10% and +10%. Only in small areas in the North and the West, the decrease is up to −20% and partly above. In the Wakhan corridor in Badakhshan and in the small areas along the border to Pakistan, stronger precipitation increases are shown by the reanalysis.
Changes in HWMI are similar to mean annual temperature in northwest-southeast oriented bands. The highest increases of over 1 of the standardized index are located in the arid areas in the southeast. The increases in the North, Centre and South regions are significant for the period 1951–2010.
Heavy precipitation between March and September increased in the East along the border to Pakistan up to 25% and above. For most of the country, however, the reanalysis shows a decrease mainly in the North, the Central Highlands and the Southwest. The changes over the whole period from 1951 to 2010 are not significant.
In
Figure 5e–g indices are plotted that are directly relevant for agriculture in Afghanistan. The SPEI indicates a decrease for most of Afghanistan, meaning an increase of drought frequency and magnitude in the past. The strongest changes are indicated for the northeastern part with a decrease of over 0.5, meaning more than half of the standard deviation. In the East smaller areas occur with an increase of SPEI, meaning a reduction of droughts. Still, none of the SPEI changes for the whole region is significant in the reanalysis data.
Spring precipitation (January to April) decreased in the northern regions and the Central Highlands of Afghanistan whereas for the southeastern part an increase is indicated. Again, averaged over the regions, the trends are not significant.
Significant GSL changes in Afghanistan extend especially along a northwest–southeast band, similar to the strongest heating. Here up to over 15 days of increase for the growing season is indicated by the reanalysis. For all of Afghanistan, from 1951 to 2010 the GSL significantly increased by over 12 days, even though it increased only by 2.8 days in the eastern region.
Changes in seasonality for monthly temperature and precipitation are depicted in
Figure 6. With regard to temperature, there is no change in seasonality visible. Temperature increased rather uniformly throughout the whole year with lowest increases at the beginning of the calendar year.
Precipitation decreased for all Afghanistan and all regions, except for the South, from March to May, and increased slightly from around June to February. This implies a slight shift of the seasons towards an earlier onset of precipitations. This pattern is most apparent in the North and Centre. In the South, no general pattern can be described but an increase in March and, less distinct, summer monsoon precipitation in July.
3.4. Analysis of Future Climate Trends for the Time Period 2006–2050/2099
The future climate model projections are, similarly to the reanalysis, analyzed in two ways: the differences between the mean of a future period 2021/2050 and a base period 1976/2005 are mapped for the different parameters (
Figure 7). In addition, linear trends are detected for the period 2006–2050 and 2006–2099 (
Table 5).
Mean annual temperature is projected to increase significantly when comparing the mean for the future period with the base period for all of Afghanistan and for all regions for the low emission scenario RCP 4.5 as well as for business-as-usual scenario RCP 8.5. For RCP 4.5, the strongest warming is in the Wakhan corridor with over 2 °C, followed by the Central Highlands with a warming of 1.75° to 2 °C. The other parts are projected to experience an increase of 1.5° to 1.7 °C. Under RCP 8.5, the spatial pattern is almost identical, however, with a magnitude that is generally 0.25 °C larger. Temperatures after 2050 until 2099 continue to increase under RCP 4.5 but less quickly. Under RCP 8.5 the warming, however, is projected to accelerate and reaches approximately 6 °C in most regions and in the Hindu Kush even over 7 °C until 2099 (
Figure 8). The model spread and the related uncertainty for the models is approximately 5 °C with all models showing very similar degrees of relative warming. The ensemble mean is more or less in the middle of the ensemble spread, indicating a homogeneous distribution.
The changes of annual precipitation are small, ranging between −5% and + 10%. Only under RCP 8.5 is there a slight wetting trend of 5–10% projected for the South, the East, and the Centre by more than 66% of the models. Until 2050, the trends are not significant with the exception of RCP 8.5 in the North. Until 2099, the signal becomes clearer and most models indicate a decrease, which is significant under RCP 8.5 for the East, North, and the Centre. The model spread is large in relation to the average, underlining the large uncertainty of the precipitation projections.
HWMI is projected to increase by all models under both scenarios for all of Afghanistan. The mean increase over all models for the future period 2021/2050 is between 1 and over 4. Given the definition of category 4–8 as extreme heat wave, this increase is substantial. The strongest increases are in the Wakhan corridor, Hindu Kush, and a large hotspot of HWMI increase covers the Centre, South and East with increases of 2–3 under RCP 4.5 and 2–4 under RCP 8.5. Until 2099, the projected increases under RCP 8.5 are enormous with +8.5 for all of Afghanistan and even 12 for the Hindu Kush region alone.
For heavy precipitation between March and September, the projected signals from the models are spatially incoherent, similar to annual precipitation. In the East and the Central Highlands, the ensemble mean indicates an increase by up to 10%. For the other parts of the country slight decreases are projected. Under RCP 4.5 there is almost no agreement among the models in terms of the trend and for RCP 8.5 over 66% of the models agree for the projected increase in the Centre. The regional trends until 2099 are similar but with even higher uncertainties and none of the regional trends of the model ensemble is significant.
The results of the analysis of the agriculturally relevant parameters are presented in
Figure 9 and
Table 6. The SPEI difference maps show a strong decrease in southern Afghanistan by over 2 standard deviations, meaning a strong increase in drought. In the other parts of the country, the SPEI is projected to decrease mainly by 1 to 0.5, and slightly below. All trends are significant but for the Hindu Kush. The differences between the scenarios until 2050 are small. The trends until 2090, however, reveal an intensification of droughts under RCP 8.5 by −2.32 for the mean of all Afghanistan. All trends under the RCP 4.5 scenario are also significant. For the future, the pattern of change of SPEI does not agree with the patterns of temperature or precipitation change. The non-linear increasing evaporation with increasing temperature can explain this effect. The smaller increase in temperature in the hot South and West of Afghanistan has a stronger effect on the evaporation (due to high absolute temperature values) compared the stronger increase in the central highlands, where absolute temperatures are, however, still lower.
The projected changes for spring precipitation are rather small between −5 to +10% and the patterns are heterogeneous. The models hardly agree in the direction of the trends, only in the North there is some agreement with respect to a small decrease of desertification in spring. The differences between the scenarios are small. Until 2099, most models show a decrease in precipitation which is significant under the RCP 8.5 scenario in all regions and in the mean for Afghanistan this decrease amounts to more than 30%. Still, the model spread is large and some models project an increase of spring precipitation.
In contrast, the signal for GSL is very obvious and significant for all regions under both scenarios and time periods. The growing season across all regions of Afghanistan, especially the central part, is projected to increase by 7 days until 2050 compared to the base period. Under RCP 8.5, the trends in all regions until 2099 are especially extreme and for the whole of Afghanistan the ensemble mean projects an increase of nearly 60 days.
The projected changes in the seasonality of temperature and precipitation are illustrated in the climate diagrams in
Figure 10 for the period from 2021 and 2050 and in the
supplementary material in Figure S2 for the future period 2070–2099. The projected temperature increase is, similar to the observations, homogeneous throughout the year. With regard to precipitation, there is a slight shift to an earlier onset of the rainy season for all regions, which is more pronounced under RCP 8.5 than under RCP 4.5.
4. Discussion and Conclusions
4.1. Robustness and Uncertainties of the Results
The validation using the 9-point filter or other interpolation-to-point techniques are inherently problematic. In this study, the area of the compared data of the nine grid cells is the mean of around 24,000 km
2. The regional geographical setting of the weather station, including its altitude, is therefore not reflected adequately. The general physical patterns should be reproduced by the reanalysis but the magnitudes of local temperature and precipitation might differ, e.g., due to topographic characteristics of the weather station. For example, Faizabad is located in the Kokcha valley (
Figure 2). Due to this lower location, temperatures are distinctly warmer than the mean of the surrounding mountainous area. Therefore, the temperature difference between the reanalysis and the measurements vary strongly but still show the same seasonal cycle. For stations in less orographic terrain, such as Kandahar, Herat and Gardez, the validation shows a more adequate performance, observed and reanalyzed temperatures being almost equal. This means that in terms of temperature, the reanalysis data at its given scale is quite reliable.
In terms of precipitation, regional differences are larger and the results are more heterogeneous, e.g., throughout the seasonal cycle. For the stations in mountainous areas, the described effect of the 9-point-filter is apparent. In addition, climate and weather models still have difficulties with high altitudes, since the spatial resolution of the cells is often too coarse to represent relevant processes that lead to rainfall at station locations. This bias is stable during the whole period. As mainly relative changes are considered for the analysis of precipitation, there is also some agreement between reanalysis and station data. Especially heavy precipitation from the reanalysis could not be validated with station data because the latter are marked by large gap.
For the regional climate model data, the same difficulties apply as for the reanalysis and additional sources of uncertainty come along. Nevertheless, the simulated seasonal cycle of temperature seems reliable. This holds for all derived parameters and indices, i.e., annual mean temperature, HWMI, SPEI, and GSL. In contrast, precipitation and the related indices that are mainly based on precipitation (i.e., annual precipitation, heavy precipitation and spring precipitation) are partly not satisfactorily reproduced by the models. In the East, the North, and the Hindu Kush, the general patterns are well represented. In contrast, the precipitation validation in the Centre and the South are, especially during spring, less promising: the models are not able to reproduce the basic patterns and, therefore, the results for these regions should not be included in the interpretation and discussion.
In summary this means that temperature and related indices (mean annual temperature, SPEI, HWMI, GSL), as well as monthly precipitation and related parameters (annual precipitation, spring precipitation) are more reliable from reanalyses compared to the sparse station data. In contrast, the performance of the GSWP3 in terms of precipitation extremes reanalysis could not be assessed, since no daily data for the validation was available. Concerning the CORDEX-SA simulations, temperature and the related indices again are reliable. For precipitation, the results of annual and spring precipitation are satisfactory in the East and North, but they are not in the Centre and South of Afghanistan.
This partly discouraging result implies a strong need for an improvement of current observations, including the potential rescue and restoration of currently unused data. It is likely that there are more analogue observations of weather stations which are currently not available due to malfunctioning administration in the responsible national ministries/agencies. With a relatively small effort, large improvements in regard of a better understanding of past climate variations and changes could be achieved, which would also help to improve current models for the future. Another option might be the exploitation of existing remote sensing data. In this regard international research has treated Afghanistan poorly and there is still a huge potential for research.
4.2. Climate Impacts
The observed and projected climatic changes found in this study tell a fairly coherent story, although there is large uncertainty with regard to precipitation. In the past, the changes are mainly characterized by temperature increases and, no matter in which direction precipitation goes, temperature increases will certainly affect ecosystems and livelihood in Afghanistan. These changes have already begun and Afghanistan has experienced a temperature increase substantially higher than in the global mean, amounting to 1.8 °C between 1951 and 2010. This warming is projected to continue from 2006 until 2050 by 1.7–2.3 °C and afterwards even 2.7–6.4 °C until 2099 across the whole country. The range is determined by the amount of global emissions. This is in accordance with the overall regional results for Central Asia which, depending on the emission scenario, are projected to be 2.5 °C compared with 2 °C globally and even 6.5 °C compared with a 4 °C global warming until the end of the century [
7]. The effects of unprecedented heat waves, as indicated by the HWMI results, will directly affect the health sector in Afghanistan. Projected increases by 8 to 16 mean that Afghanistan will regularly experience very extreme heat waves, especially under the business-as-usual emission scenario RCP 8.5. Temperature will also impact on water resources and agriculture as well as on natural ecosystems in Afghanistan. Increasing evapotranspiration as indicated by the SPEI will probably exacerbate the country’s already difficult conditions.
Although Afghanistan has vast water resources (approximately 1700 m
3 per capita per year, which in theory is sufficient for domestic, agricultural, industrial, and environmental needs), the country still does not use the resources efficiently due to a lack of water management capacity in the face of the large regional as well as intra- and interannual variability of water availability [
72]. In the future, this situation may become even more problematic given the growing demand related to population growth, economic development, and improved standards of living. Despite Afghanistan’s strong dependency on agriculture, which employs around 85 percent of the population directly or indirectly and contributes around 30% to the gross domestic product [
72], only very few research and impact studies for agriculture and water resources have been conducted (see
Section 1). Only the glaciers of the Hindu Kush have attracted the attention of some international research activities [
13,
14,
16,
17,
18,
20].
The negative impacts of climate change not only hold for the agricultural sector but also for Afghanistan’s natural ecosystems, which have already been deteriorated during the country’s many years of conflict, unsustainable management, and over-exploitation. For example, over 80% of Afghanistan’s land is said to be subject to soil erosion [
36]. This discussion can also be extended for the hazard landscape of Afghanistan. The numerous regularly occurring hazards like avalanches, floods, droughts, landslides but also long-term erosion are related to the changing climate [
73]. Also in this regard, there is a large research deficit for Afghanistan. The contribution of this study to hazard research is limited due to its more general character; however, it is likely that droughts will occur more often and hazards related to runoff, such as landslides, floods and flash floods, are likely to be enhanced by climate change [
74,
75].
Overall, the current results show that in addition to the already existing deficit in adaptation to current climate conditions, the situation will be aggravated in the future, particularly in the areas of water management and agriculture. Thus, the results of this study underline the importance of adaptation to climate change in Afghanistan.
This holds even more because there is also a clear positive signal in the results. The GSL will increase substantially with rising temperatures, e.g., by around 20 days on average until 2050. This might open the opportunity for extended agricultural usage or in some cases even an additional harvest. This requires, of course, a smart management of water resources and a more sophisticated and climate-adapted agriculture.